Test MergeToolTest#testEmptyToolName started failing long time after it was implemented

[msohn]

Version

stable-6.10 (f27ea51)

Operating System

Linux/Unix

Bug description

The test which is only run on Linux succeeded many times after it was implemented in https://eclipse.gerrithub.io/c/eclipse-jgit/jgit/+/140275.

Since last week it started failing with an assertion error, see e.g. https://ci.eclipse.org/jgit/job/stable/job/jgit.gerrit-pipeline.java11/3016/consoleFull.

Maybe the failure is caused by some change in the build environment or configuration which is used on the CI system.

Actual behavior

The test fails reproducibly

Expected behavior

The test should succeed

Relevant log output

03:34:38  [ERROR] Tests run: 11, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 5.562 s <<< FAILURE! -- in org.eclipse.jgit.pgm.MergeToolTest
03:34:38  [ERROR] org.eclipse.jgit.pgm.MergeToolTest.testEmptyToolName -- Time elapsed: 0.709 s <<< FAILURE!
03:34:38  org.junit.ComparisonFailure: Expected no standard error output from tool 
expected:<[compare[: unrecognized option `-wait' @ error/compare.c/CompareImageCommand/1123., compare: unrecognized option `-wait' @ error/compare.c/CompareImageCommand/1123].]> 
but was:<[compare[-im6.q16: unrecognized option `-wait' @ error/compare.c/CompareImageCommand/1131., compare-im6.q16: unrecognized option `-wait' @ error/compare.c/CompareImageCommand/1131].]>
03:34:38  	at org.junit.Assert.assertEquals(Assert.java:117)
03:34:38  	at org.eclipse.jgit.pgm.ToolTestCase.runAndCaptureUsingInitRaw(ToolTestCase.java:112)
03:34:38  	at org.eclipse.jgit.pgm.ToolTestCase.runAndCaptureUsingInitRaw(ToolTestCase.java:79)
03:34:38  	at org.eclipse.jgit.pgm.MergeToolTest.testEmptyToolName(MergeToolTest.java:96)

Other information

No response

[msohn]

Change disabling this test to unblock other unrelated changes until the cause of this test failure is found and fixed: https://eclipse.gerrithub.io/c/eclipse-jgit/jgit/+/1205021

[tomaswolf]

The test assumes a particular stderr output from an external program. Looks like we now have a different version of that external program, which produces a slightly different stderr output.

This is most likely a result of the OS change on the Jenkins agents. Our tests now run on Ubuntu 24.04.